{smcl}
{* 10feb2009}{...}
{cmd:help mata mm_rbinomial()}
{hline}

{title:Title}

{p 4 19 2}
{bf:mm_rbinomial() -- Generate binomial distributed random numbers}

{hline}
{p 2 2 2}Note: An official {cmd:rbinomial()} function was introduced
in Stata 10.1 (see help
{bf:{help mf_runiform:[M-5] runiform()}}). Official {cmd:rbinomial()} is much faster
than {cmd:mm_rbinomial()}.
{p_end}
{hline}


{title:Syntax}

{p 8 23 2}
{it:real matrix}
{cmd:mm_rbinomial(}{it:n}{cmd:,} {it:p}{cmd:)}

{p 8 23 2}
{it:real matrix}
{cmd:mm_cebinomial(}{it:n}{cmd:,} {it:k}{cmd:,} {it:p}{cmd:)}

{pstd}
where

        {it:n}:  {it:real matrix n}
        {it:k}:  {it:real matrix k}
        {it:p}:  {it:real matrix p}


{title:Description}

{pstd}{cmd:mm_rbinomial()} returns a random draw from the binomial
distribution where {it:n} is the number of trials and {it:p}
is the success probability.

{pstd}
When {it:n} and {it:p} are not scalar, {cmd:mm_rbinomial()}
returns element-by-element results. {it:n} and {it:p} are
required to be r-conformable (see help
{helpb m6_glossary:[M-6] glossary}).

{pstd}If you are serious about generating random numbers,
you should first set the random number seed; see help {helpb generate}
or help for {helpb mf_uniform:[M-5] uniform()}.

{pstd}{cmd:mm_cebinomial()} returns the expected value of a binomial
distributed random variable
conditional on the variable being equal to
{it:k} or larger. That is, {cmd:mm_cebinomial()} returns

{phang2}
E({it:X}|{it:X}>={it:k})

{pstd}
where

{phang2}
{it:X} ~ B({it:n}, {it:p})

{pstd}
and {it:n} is the number of trials and {it:p} is the success
probability.

{pstd}
When {it:n}, {it:k}, and {it:p} are not scalar, {cmd:mm_cebinomial()}
returns element-by-element results. {it:n}, {it:k}, and {it:p} are
required to be r-conformable (see help
{helpb m6_glossary:[M-6] glossary}).


{title:Remarks}

{pstd}{cmd:mm_rbinomial()}: If {it:n}<50 or {it:p}>.03, the random numbers are generated using
the rejection technique. If {it:n}>=50 and {it:p}>=.03, the geometric distribution
technique is applied (see Evans et al. 2000).

{pstd}{cmd:mm_cebinomial()}: The expectation of {it:X} ~ B({it:n}, {it:p}) conditional on
{it:X}>={it:k} may be written as

{phang2}
E({it:X}|{it:X}>={it:k}) = {it:k} + [ P({it:X}>={it:k}+1) + ... + P({it:X}={it:n}) ] / P({it:X}>={it:k})

{pstd}
where P({it:X}>={it:k}) is the probability of {it:k} or more
successes, which is computed as {opt Binomial(n, k, p)}
(see {helpb mf_normal:[M-5] normal()}).


{title:Conformability}

{pstd} Both functions require arguments be r-conformable (see help
{helpb m6_glossary:[M-6] glossary}). Returned is a matrix of
max(argument rows) rows and max(argument columns)
columns containing element-by-element calculated results.


{title:Diagnostics}

{pstd}
Missing value is returned if arguments are out of range
({it:p}>1, {it:p}<0, {it:n}<=0, {it:k}<0, or {it:k}>{it:n})
or contain missing value or if {it:n} and {it:k} are non-integer.


{title:Source code}

{pstd}
{help moremata_source##mm_rbinomial:mm_rbinomial.mata},
{help moremata_source##mm_cebinomial:mm_cebinomial.mata}


{title:References}

{phang}
Evans, M., N. Hastings, B. Peacock (2000): Statistical Distributions.
Third Edition. New York: Wiley.


{title:Author}

{pstd}
Ben Jann, University of Bern, jann@soz.unibe.ch


{title:Also see}

{psee}
Online:  help for {bf:{help mf_normal:[M-5] normal()}},
{bf:{help moremata}}

{psee}
Links to user-written programs:
{net "describe genbinomial, from(http://www.stata.com/users/rgutierrez)":genbinomial},
{net stb 41 sg44.1:sg44.1}
